002_分布式搜索引擎Elasticsearch的查询与过滤

一、写入

先来一个简单的官方例子,插入的参数为-XPUT,插入一条记录。

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
curl -XPUT 'http://localhost:9200/test/users/1' -d '{
    "user""test",
    "post_date""2009-11-15T14:12:12",
    "message""Elastic Search"
}'
 
{
    "_index""test",
    "_type""users",
    "_id""1",
    "_version": 2,
    "_shards": {
        "total": 2,
        "successful": 1,
        "failed": 0
    },
    "created"false
}

从上面的这个例子中,可以看出ES的http的服务的默认端口9200,后面的/test/users/1是这条记录的索引部分,体现了它的RESTful风格

这三级目录分布对应了_index,_type,_id 实际上ES上存放的所有的记录都只能通过三级目录的方式找到

  • _id 字段可以是数字也可以是字符串。在执行上面的命令时ES会自动创建这些索引

  • -d 后面跟上了要插入的json格式的记录

  • -XPUT 表明这是插入一条数据,ES中叫创建一个索引

  • _version 字段,表明了当前记录的版本号,当你想这个索引重新put一条记录时,版本号会自动加一

二、删除

删除的http请求参数为-XDELETE,通过下面的命令可以删除这条记录:

1
2
3
4
5
6
7
8
9
10
11
12
13
curl -XDELETE 'http://localhost:9200/test/users/1?pretty'
{
  "found" true,
  "_index" "test",
  "_type" "users",
  "_id" "1",
  "_version" : 3,
  "_shards" : {
    "total" : 2,
    "successful" : 1,
    "failed" : 0
  }
}

删除这条记录的时候,_verison也会自动加一的。

三、查询

创建了一个索引后,可以通过下面的方式查询(参数-XGET)出来

1
2
3
4
5
6
7
8
9
10
11
12
13
curl -XGET 'http://localhost:9200/test/users/1?pretty'      
{
  "_index" "test",
  "_type" "users",
  "_id" "1",
  "_version" : 1,
  "found" true,
  "_source" : {
    "user" "test",
    "post_date" "2009-11-15T14:12:12",
    "message" "Elastic Search"
  }
}

执行上面的查询命令,可以等到下面的结果,exists表示是否有查询结果,_source字段是查询到的记录。

查询的时候,可以将_type设置成为_all,ES就会返回在_index下所有type中,第一个匹配_id的记录,还可以通过参数对返回结果继续控制,

用fields选取返回的字段,用pretty控制返回的json格式是否更阅读友好

1
2
3
4
5
6
7
8
9
10
11
12
curl -XGET 'http://localhost:9200/test/users/1?fields=message,user&pretty=true'
{
  "_index" "test",
  "_type" "users",
  "_id" "1",
  "_version" : 3,
  "found" true,
  "fields" : {
    "message" : [ "Elastic Search" ],
    "user" : [ "test" ]
  }
}

format=yaml可以设置输入格式为YAML

1
2
3
4
5
6
7
8
9
10
11
12
curl -XGET 'http://localhost:9200/test/users/1?fields=message,user&format=yaml'
---
_index: "test"
_type: "users"
_id: "1"
_version: 1
found: true
fields:
  message:
  "Elastic Search"
  user:
  "test"

当然ES还支持一次查询多组记录,即multi get,在URI中是使用关键字_mget,具体可以参考ES的文档multi get

四、更新

ES同样支持更新,但是更新的方式是通过一个提供的脚本进行的。

ES的做法是,通过index找到相应的存放记录的节点,然后执行脚本,执行完之后,返回新的索引。实际上执行的是一个get和reindex的过程,在这个过程中,通过versioning来控制没有其它的更新操作(这个功能是0.19后可用的)。具体实现的原理应该和elasticsearch Versioning相关。

get,reindex的含义是,ES先取出这条记录,然后根据新数据生成新记录,然后在把新记录放回到ES中(并不会覆盖老的记录)。

首先创建一条记录

1
2
3
4
curl -XPUT localhost:9200/test/type1/1 -d '{
    "counter" : 1,
    "tags" : ["red"]
}'

将counter的值加4

1
2
3
4
5
6
curl -XPOST 'localhost:9200/test/type1/1/_update' -d '{
    "script" "ctx._source.counter += count",
    "params" : {
        "count" : 4
    }
}'

也可以添加一个tag的值

1
2
3
4
5
6
curl -XPOST 'localhost:9200/test/type1/1/_update' -d '{
    "script" "ctx._source.tags += tag",
    "params" : {
        "tag" "blue"
    }
}'

现在还支持upsert功能,即在更新的时候,如果记录没有这个key,则插入这个key,下面是一个例子,如果没有counter字段,则插入该字段:

1
2
3
4
5
6
7
8
9
curl -XPOST 'localhost:9200/test/type1/1/_update' -d '{
    "script" "ctx._source.counter += count",
    "params" : {
        "count" : 4
    },
    "upsert" : {
        "counter" : 1
    }
}'

关于update还有其它很多功能,可以参考ES的API update

五、搜索

elasticsearch的名字里面有一个search,那么主要功能也是search了。

es的search有两种形式,一是通过URI,二是通过Requst Body。通过URI查询,即将查询的语句放入到请求的url中,例如:

1
curl -XGET 'http://localhost:9200/twitter/tweet/_search?q=user:kimchy'

第二种方式,即在查询的请求中加入一个doc

1
2
3
4
5
curl -XGET 'http://localhost:9200/twitter/tweet/_search' -d '{
    "query" : {
        "term" : { "user" "kimchy" }
    }
}'

query body的定义可以查看query DSL 另外两种查询方式都可以带参数,参数的含义参考URI Request和Request Body

ES的搜索功能是可以跨index和type的,例如下面这几条命令

1
2
3
4
5
curl -XGET 'http://localhost:9200/twitter/_search?q=user:kimchy'
curl -XGET 'http://localhost:9200/twitter/tweet,user/_search?q=user:kimchy'
curl -XGET 'http://localhost:9200/kimchy,elasticsearch/tweet/_search?q=tag:wow'
curl -XGET 'http://localhost:9200/_all/tweet/\_search?q=tag:wow'
curl -XGET 'http://localhost:9200/\_search?q=tag:wow'

第一条是在所有的twitter这个index下的所有type中查找,第二条是在tweet,user这两个type中查找,第三条是在kimchy,elasticsearch这两个index的tweet这个type中查找,第四条使用了_all关键字,是在所有的index的tweet的type中查找,第五条更暴力,在所有的index和type中查找。

查找还有其它的很多选项,sort高亮,选取返回记录的域Fields,还可以对返回的域使用一个脚本进行计算script Fields,或者对返回结果继续统计Facets,Facets的内容比较多,它支持关键词统计,范围内统计,直方图式统计,日期的直方图式统计,过滤,查询,还有记录地理位置距离的统计geo distance。 支持名字过滤Named Filters。 定义搜索类型Search Type 。例如什么Query And Fetch,Query Then Fetch。 索引加速的功能Index Boost,可以让某一个索引的权重大于另外一个。 保持上次检索的环境了结果Scroll。保留每一个命中的score值Explain。 设置命中的min_score。保留版本号Version

Search的参数很多,我也没有一一看,不过果然是名字里面有个search,对检索的各种场景都有支持。

当然还支持多个查询multi search,例如下面这个例子

1
2
3
4
5
6
7
8
9
cat requests
{"index" "test"}
{"query" : {"match_all" : {}}, "from" : 0, "size" : 10}
{"index" "test""search_type" "count"}
{"query" : {"match_all" : {}}}
{}
{"query" : {"match_all" : {}}}
  
$ curl -XGET localhost:9200/_msearch --data-binary @requests; echo

六、DSL

1、match

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
curl -XGET 'localhost:9200/test/_search?pretty' -d'{
    "_source": ["title""allnum"],
    "query": {
        "match": { "title""地铁跑酷" }
    }
}'
 
{
  "took" : 2,
  "timed_out" false,
  "_shards" : {
    "total" : 5,
    "successful" : 5,
    "failed" : 0
  },
  "hits" : {
    "total" : 1,
    "max_score" : 6.692047,
    "hits" : [ {
      "_index" "test",
      "_type" "external",
      "_id" "AVicbl1W9iJMapPz5wea",
      "_score" : 6.692047,
      "_source" : {
        "title" "地铁跑酷",
        "allnum" : 104541
      }
    } ]
  }
} 

显示指定字段,匹配关键字

结果格式

  • took – 搜索用的毫秒

  • timed_out – 搜索超时时间

  • _shards – 搜索了多少个片段 成功搜索多少个 失败了多少个

  • hits – 搜索的返回结果集

  • hits.total – 结果的总数

  • hits.hits – 实际搜索返回的数组

  • _score and max_score - ignore these fields for now

一旦es搜索返回给你 该链接就断开了 es并不会想MySQL之类的 一直维护一个连接

更多限制条件的搜素

1
2
3
4
5
6
7
8
9
curl -XGET 'localhost:9200/test/_search?pretty' -d'{
    "from": 0,
    "size": 1,
    "_source": ["title""allnum"],
    "query": {
        "match": { "title""地铁跑酷" }
    },
   "sort": { "allnum": { "order""desc" }}
}'

2、使用should

1
2
3
4
5
6
7
8
9
10
11
12
13
14
curl -XGET 'localhost:9200/test/_search?pretty' -d'{
    "from": 0,
    "size": 2,
    "_source": ["title""allnum"],
    "query": {
        "bool": {
            "should": [
                "match": { "title""地铁跑酷" } },
                "match": { "title""星猫跑酷" } }
            ]
        }
    },
   "sort": { "allnum": { "order""desc" }}
}'

3、使用must/must not

1
2
3
4
5
6
7
8
9
10
11
12
13
14
curl -XGET 'localhost:9200/test/_search?pretty' -d'{
    "from": 0,
    "size": 2,
    "_source": ["title""allnum"],
    "query": {
        "bool": {
            "must": [
                "match": { "title""地铁跑酷" } },
                "match": { "title""星猫跑酷" } }
            ]
        }
    },
   "sort": { "allnum": { "order""desc" }}
}'

当然 must /should/must not 可以联合起来使用 

4、过滤器

1
2
3
4
5
6
7
8
9
10
curl -XPOST 'localhost:9200/test/_search?pretty' -d'{
    "query": {
        "filtered": {
            "query": {
                 "match_all": {}
              },
             "filter": {
                 "range": { "allumn": { "gte": 20000, "lte": 30000 } } } }
               }
}'

5、聚合

1
2
3
4
5
6
7
curl  -XPOST  'localhost:9200/test/_search?pretty' -d'{
    "size": 0,
    "aggs": {
        "group_by_state": {
            "terms": { "field""state" } }
         }
}'

6、terms 过滤

1
2
3
4
5
6
7
8
9
10
curl  -XPOST  'localhost:9200/test/_search?pretty' -d'{
  "query": {
    "terms": {
      "status": [
        304,
        302
      ]
    }
  }
}

7、range 过滤

1
2
3
4
5
6
7
8
9
curl  -XPOST  'localhost:9200/test/_search?pretty' -d'{
  "query": {
    "range": {
      "allnum": {
        "gt": 1
      }
    }
  }
}

范围操作符包含:

  • gt :: 大于

  • gte:: 大于等于

  • lt :: 小于

  • lte:: 小于等于

8、exists 和 missing 过滤

exists 和 missing 过滤可以用于查找文档中是否包含指定字段或没有某个字段,类似于SQL语句中的IS_NULL条件

1
2
3
4
5
{
    "exists":   {
        "field":    "title"
    }
}

9、bool 过滤

bool 过滤可以用来合并多个过滤条件查询结果的布尔逻辑,它包含一下操作符:

  • must :: 多个查询条件的完全匹配,相当于 and。
  • must_not :: 多个查询条件的相反匹配,相当于 not。
  • should :: 至少有一个查询条件匹配, 相当于 or。

这些参数可以分别继承一个过滤条件或者一个过滤条件的数组:

1
2
3
4
5
6
7
8
9
10
{
    "bool": {
        "must":     { "term": { "folder""inbox" }},
        "must_not": { "term": { "tag":    "spam"  }},
        "should": [
                    "term": { "starred"true   }},
                    "term": { "unread":  true   }}
        ]
    }
}

七、中文分词

创建索引

1
curl -XPUT http://localhost:9200/index

创建映射

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
curl -XPOST http://localhost:9200/index/fulltext/_mapping -d'
{
    "fulltext": {
             "_all": {
            "analyzer""ik_max_word",
            "search_analyzer""ik_max_word",
            "term_vector""no",
            "store""false"
        },
        "properties": {
            "content": {
                "type""string",
                "store""no",
                "term_vector""with_positions_offsets",
                "analyzer""ik_max_word",
                "search_analyzer""ik_max_word",
                "include_in_all""true",
                "boost": 8
            }
        }
    }
}'

为索引添加一些内容

1
2
3
4
5
6
7
8
9
10
11
12
curl -XPOST http://localhost:9200/index/fulltext/1 -d'
{"content":"美国留给伊拉克的是个烂摊子吗"}
'
curl -XPOST http://localhost:9200/index/fulltext/2 -d'
{"content":"公安部:各地校车将享最高路权"}
'
curl -XPOST http://localhost:9200/index/fulltext/3 -d'
{"content":"中韩渔警冲突调查:韩警平均每天扣1艘中国渔船"}
'
curl -XPOST http://localhost:9200/index/fulltext/4 -d'
{"content":"中国驻洛杉矶领事馆遭亚裔男子枪击 嫌犯已自首"}
'

进行查询

1
2
3
4
5
6
7
8
9
10
11
12
curl -XPOST http://localhost:9200/index/fulltext/_search  -d'
{
    "query" : { "term" : { "content" "中国" }},
    "highlight" : {
        "pre_tags" : ["<tag1>""<tag2>"],
        "post_tags" : ["</tag1>""</tag2>"],
        "fields" : {
            "content" : {}
        }
    }
}
'  

参考:http://www.cnblogs.com/chenpingzhao/p/6067350.html

原文地址:https://www.cnblogs.com/itcomputer/p/6751422.html